Nice polishing job, Kawi!
My mod yesterday, just a few pics of the mod:
I had this triple for some time, waiting for an idea:
From Aliexpress: link
Cheaply made, it came with chinese lettering that is chinese to me, but with an interesting driver: a toggle between the three leds, with on the (20mm diameter) driver for each channel a little FET and a 1.2 Ohm resistor in series for current limiting. So this driver can do more! The design does not support that though, a P60-like floating pill pushed against the lens with a spring: the heat management is not great.
I decided to not make use of the driver (it sits in the parts box now for a future mod), but make use of the triple reflector feature: UV-leds can not use plastic TIR optics because it absorbs the UV, so a UV-led triple must either be build as a mule or use a triple reflector.
What did I do?
*After precisely measuring what length was needed, I replaced the spring by a copper ring. One of the EU standard copper piping sizes is exactly the size and nicely fits an 18650 inside. This copper ring should transfer the heat from the pill well to the (pretty thick) battery tube.
*The three plus-pads on the MCPCB were connected by a (sorry about the wrong colour) copper wire to make it a triple parallel board
*Three Lite-On 365nm 3535 leds were reflowed on the board (my test of this led here)
*the three (connected) plastic led centerpieces were opened up a bit more with a countersink tool to use the reflector more effectively
*The conical (made for gem-inspection) bezel was sawed off and sanded flat to make it into a (still pretty deep) standard type bezel
*the ZWB2 filter was too small to fit the inner diameter of the head, but just wide enough to cover the bezel opening, so it was placed on top of the reflector with a black o-ring pressed snuggly around it to center it (and waterproofing the head as well, the stock light had no waterproofing there)
*I wanted a 8x 7135 driver, to get 1A for each led, not more because it is not a DTP board. But I have no 20mm lineair drivers so I had to use a brass ring soldered onto a 17mm driver to get it fit, but that ring did not easily allow the 7135-chips on the spring-side, and the pill was too shallow to stack the 4 chips on the led-side of the driver. So I gambled on a direct driver and hoped that the relatively high voltage of these leds, the resitance of the springs and the use of a low-drain battery would keep the current in control 
*swapped the unknown quality switch for a small Omten on a 16mm PCB, this is an almost direct fit. Replaced the red tailcap for a blue one which had the correct depth for the new switch.
Just before closing up:
Aaaaand gone:
The bezel pushes directly against the filter that via the reflector presses the ledboard to the pill and the pill to the copper ring. So the complete assembly is pressed tight without the dampening of any o-ring, maybe a tiny bit by the centering pieces. Good for heatsinking and if the filter breaks at some point I have enough spares.
Then I got careless and tried the first battery I could find: a NCR18650PF, so pretty high drain. I measured more than 6 amps, not good for the leds, 
. And sure enough:
So it was repairing time! Opening it all up again, replace the damaged led (the bond wires were visibly blown), replacing the led wires from driver to ledboard with the thinnest ones I could find (must be 30AWG or so) and assembling again. Now, on a NCR18650B (a lot lower drain cell) I measured, without tail, 3.7 A, so with tail on a bit less, maybe 3.5A. That feels healthy enough for these leds.
Here's projection (projection lens before the flashlight) of the leds on a piece of white paper:
It looks like the remaining two leds survived the 2+ amps without visibly loosing too much output :-)
It was not easy to catch the beam, here's an attempt on a white wall, the center of the hotspot looks messed-up in the picture but it is fine in reality. A nice broad hotspot and plenty spill.
And it is bright alright. Meet my gas stove after I prepared the evening meal yesterday:
You did not want to see that! :party:
So now I'm stuck with the the brightest 365nm light that I made to date. I do not have a way to measure radiometric output but it must be over 2W output. I know no one who works in forensics, but I would guess that this brings the 365nm light that they normally use to a new level
A few glampics:
Disclaimer: this light puts out a truck-load of completely invisible UV light that if watched directly or even in diffused reflection will (possibly permanently) damage your eyes. May you repeat such a mod, wear safety goggles (standard polycarbonate goggles will block 99% of the 365nm UV light, but test your goggles first to be certain) and do not use the light for a prolonged period.
